Auto_Jobs_Applier_AI_Agent by AIHawk is an AI Agent that automates the jobs application process. Utilizing artificial intelligence, it enables users to apply for multiple jobs in an automated and personalized way.
Logic to skip an input text-box if the field is already populated and only answer required questions
Feature description
Overall, these ideas aim to reduce the impact of hallucinations.
I successfully used the bot to apply to ~100 jobs, however after reviewing the applications I don't believe these applications will bear any fruit as the system hallucinated quite often.
Sometimes the hallucinations were funny, such as making up my gender identity or stating that I had completed a PhD.
I was most concerned with how often the bot failed to answer work authorization questions.
Feature 1
LinkedIn auto-populates certain questions, such as first and last name, email, address, phone number, and even job-specific questions such as "How many years of Python do you have?"
If the system could identify that these questions were already populated, hallucinations wouldn't matter.
Feature 2
The next feature pertains to the required questions.
Instead of answering all the questions, would it be possible to answer only those questions marked with an asterisk?
When I apply manually I skip these questions anyway.
Motivation
~100 applications but with wrong/false info probably won't lead to interviews
Alternatives considered
Hot keys.
The bot took 5 hours to apply to ~100 jobs.
I can accomplish the same thing in < 2 hours using hotkeys and skipping optional questions.
Most importantly though, I don't hallucinate (usually ;) lol)
Feature summary
Logic to skip an input text-box if the field is already populated and only answer required questions
Feature description
Overall, these ideas aim to reduce the impact of hallucinations. I successfully used the bot to apply to ~100 jobs, however after reviewing the applications I don't believe these applications will bear any fruit as the system hallucinated quite often. Sometimes the hallucinations were funny, such as making up my gender identity or stating that I had completed a PhD. I was most concerned with how often the bot failed to answer work authorization questions.
Feature 1 LinkedIn auto-populates certain questions, such as first and last name, email, address, phone number, and even job-specific questions such as "How many years of Python do you have?" If the system could identify that these questions were already populated, hallucinations wouldn't matter.
Feature 2 The next feature pertains to the required questions. Instead of answering all the questions, would it be possible to answer only those questions marked with an asterisk? When I apply manually I skip these questions anyway.
Motivation
~100 applications but with wrong/false info probably won't lead to interviews
Alternatives considered
Hot keys. The bot took 5 hours to apply to ~100 jobs. I can accomplish the same thing in < 2 hours using hotkeys and skipping optional questions. Most importantly though, I don't hallucinate (usually ;) lol)
Additional context
Model: llama3.2:3b via Ollama on docker using GPU